rdb_config Module

Initial salinity at the bed (k=1, PSU).

Initial salinity at the surface layer (k=nz_ml, PSU). Activates linear S(z) stratification when set non-zero together with S_init_bottom.

Upper physical bound for salinity (PSU)

Lower physical bound for salinity (PSU)

RETIRED coastal-legacy EOS reference salinity (PSU). The live ocean-path spelling is &ocean_ic_nml S_ref; moving this one off its default is a fail-loud configure error.

Initial temperature at the bed (k=1, degC).

Initial temperature at the surface layer (k=nz_ml, degC). Activates linear T(z) stratification when set non-zero together with T_init_bottom.

Upper physical bound for temperature (degC)

Lower physical bound for temperature (degC); seawater freezing

RETIRED coastal-legacy EOS reference temperature (degC). The live ocean-path spelling is &ocean_ic_nml T_ref; moving this one off its default is a fail-loud configure error.

RETIRED coastal-legacy thermal expansion coefficient (kg/m^3 per degC). Live spelling: &ocean_ic_nml alpha_T.

Path to NetCDF bathymetry file (empty = flat bottom)

East boundary: “wall”, “open”, “tidal”, “nested”

North boundary: “wall”, “open”, “tidal”, “nested”

South boundary: “wall”, “open”, “tidal”, “nested”

West boundary: “wall”, “open”, “tidal”, “nested”

RETIRED coastal-legacy haline contraction coefficient (kg/m^3 per PSU). Live spelling: &ocean_ic_nml beta_S.

CFL number for adaptive timestep

Recompute CFL timestep every N steps (1 = every step)

I1′ tripwire — assert h_layer <= H_VANISHED ⇒ hTr = h_layer·c_live (the donor live layer’s concentration; hTr = 0 in a column with no live layer) for every registered tracer, once per outer step, immediately after the enforcement point that establishes it (multilayer_state_t%enforce_vanished_content). A violation logs the offending cell count and the worst |hTr − h·c_live| and error stops.

Deflate compression level (1=fast, 9=max). 1 is usually optimal.

Enable deflate compression for NetCDF output (NetCDF4/HDF5)

Coriolis parameter f (1/s), typically 2Omegasin(lat)

Fixed timestep (s), 0 = adaptive CFL. Same convention as dt_max — always seconds.

Maximum allowable timestep (s) — always in seconds, NOT affected by time_unit. Timesteps are naturally short and CFL-bounded; writing them in "day" would force awkward fractional values.

Cell size in x-direction (m)

Cell size in y-direction (m)

Background depth for gaussian_hump IC (m)

Horizontal diffusion coefficient for tracers (m^2/s). VESTIGIAL: its only consumers were the coastal solvers, which left with the carve-out, so nothing reads it today. The along-coordinate equivalent is the separate &ocean_hdiff_nml kappa_h knob (rdb_ocean_hdiff_tracer). Smooths sharp density fronts to reduce BPG overshoot. Typical values: 1-10 m^2/s depending on grid resolution.

Prescribed salinity for inflow BC (PSU); <0 = zero-gradient

Prescribed temperature for inflow BC (degC); <0 = zero-gradient

Initial salinity for all layers (PSU). Used as a uniform IC unless S_init_surface and S_init_bottom are BOTH set non-zero, in which case the seed builds a linear S(z) profile from S_init_bottom at k=1 (bed) to S_init_surface at k=nz (surface). NOTE the stable polarity is the inverse of temperature: salty/dense water belongs at the bed, so a stable haline column has S_init_bottom > S_init_surface (fresh surface, e.g. a river-plume column).

Initial potential temperature for all layers (degC). Used as a uniform IC unless T_init_surface and T_init_bottom are BOTH set non-zero, in which case the seed builds a linear T(z) profile from T_init_bottom at k=1 (bed) to T_init_surface at k=nz (surface). Stratified IC is the prerequisite for baroclinic-instability-driven eddies.

Background vertical salinity diffusivity (m^2/s)

Background vertical temperature diffusivity (m^2/s)

Unresolved-turbulence coefficient for the V_t^2 term in the bulk-Ri denominator (LMD94 eq 23). Set 0 to disable V_t^2.

Non-local (counter-gradient) transport coefficient C_s (LMD94 eq 20, limit value).

Critical bulk Richardson number for the KPP BL-depth sweep.

Log verbosity: “debug”, “verbose”, “info”, “performance”, “warning”, “error”

Manning roughness coefficient

Number of active tidal constituents (0 = use legacy single)

Number of ghost cells on each side. Default 3 (not 2): the ocean dyn-core’s continuity/tracer PPM reconstruction is a 5-point stencil, so a rank-seam face’s ghost-side donor needs two neighbours beyond itself — full-order reconstruction AT the seam requires three ghost columns. At nghost=2 the PPM local-array-edge fallback fires at the seam and degrades the face to first order, producing a seam-inconsistent tracer mass flux: it conserves a uniform tracer (constant field cancels) but LEAKS a structured one (advecting T with rank-A-out /= rank-B-in on the shared seam face drifts hT), invisible to the mass/salt closure gates. Measured on seamount_bench_full (np2 x-split): heat closure grew to +5e-11 by day 2 at nghost=2, stays at round-off (~5e-14, no growth) at nghost=3. Single-rank runs are unaffected by the wider halo (same interior); the extra ghost column only costs a little memory.

Number of physical cells in x-direction

Number of physical cells in y-direction

Number of sigma layers for NH solver

Nested ocean knobs. Per-concern sub-types live under here (coriolis, thermo, bt, pgf, bdrag, hvisc, vmix, continuity, topo, ic, diag). PR-B1 migrates knobs from this struct’s flat fields into the nested form one group at a time — callers go from cfg%ocean_X → cfg%ocean%group%X. PR-B2 will split the giant &ocean_setup_nml to match.

Directory for output files

Enable file output (NetCDF snapshots)

Number of MPI processes in x-direction

Number of MPI processes in y-direction

Grid time-filter timescale τ (s). The ALE regrid relaxes the coordinate a fraction dt/(τ+dt) toward the target each thermo step instead of jumping (White & Adcroft 2008), damping the σ/z* per-step grid-motion shock. 0 (default) = jump to target = bit-identical.

Linear-exact one-sided reconstruction in the ALE remap’s two boundary cells (MOM6 BOUNDARY_EXTRAPOLATION) instead of the PCM flatten, which leaves PLM/PPM/PPM_H4/PQM first-order at k=1 and k=nz. .false. (default) = bit-identical.

Assert, once per ALE remap (thermo cadence), that every column satisfies what the overlap sweep has always assumed: non-negative source AND target thicknesses, and matching column totals. Both are caller obligations and neither was ever checked; a violation silently CREATES or DELETES tracer mass. Diagnostic knob — .false. (default) = the check never runs = bit-identical.

Vertical remapping method: “pcm”, “plm”, “ppm”, “ppm_h4”, “pqm”. Default PPM (piecewise parabolic, Colella & Woodward 1984) — the state-of-the-art conservative vertical-remap stencil for the coastal ALE coords. “pqm” is piecewise quartic (White & Adcroft 2008); it falls back to PPM for nz < 5.

Non-uniform-grid reconstruction weights in the ALE remap’s PLM slope and PPM edge estimate (Colella & Woodward 1984 eqs 1.6-1.8) instead of their equal-thickness specialisations, which are linear-exact only on a UNIFORM source column. PPM_H4 and PQM already carry thickness-weighted stencils and are unchanged. .false. (default) = bit-identical.

KE-conserving rescale of the remapped layer velocities: scale the baroclinic anomaly per column so column KE is preserved (Adcroft & Hallberg 2006), capped 1.25×; barotropic mean untouched. .false. (default) = momentum-only = bit-identical.

Path to restart file for warm start (empty = cold start)

Time between restart file writes (s), 0 = no restarts

Reference density for EOS (kg/m^3)

Reference pressure (Pa, default 2e7 = 2000 dbar) for the potential density that defines the VCOORD_RHO coordinate.

Densest target interface potential density (kg/m³) — the bed interface for VCOORD_RHO / VCOORD_HYCOM. See rho_target_light.

Lightest target interface potential density (kg/m³) — the surface interface for VCOORD_RHO / VCOORD_HYCOM. With rho_target_dense builds the uniform light→dense target-density linspace (MOM6 ALE_COORDINATE_CONFIG=UNIFORM analogue) under rho_target_profile = "uniform".

rho_target_profile = "list": interface potential densities (kg/m³) at rho_ref_pressure, LIGHTEST (surface) FIRST, strictly increasing. Exactly nz_layers+1 leading positive entries; the rest unset (<= 0, default -1).

Target-density profile of vcoord_type = "rho" | "hycom": "uniform" (default — the rho_target_light→rho_target_dense linspace, byte-identical) or "list" (the nz_layers+1 interface densities in rho_target_list, light first — MOM6 target densities from a list/file, e.g. the HYBRID:file,sigma2,dz coordinate of OM4). A uniform list is a poor fit wherever most of the volume sits in a narrow density range (the Southern Ocean: half its volume inside ~2 linspace layers). Refused on any other coordinate.

Simulation regime: “coastal” or “ocean”

Sponge relaxation rate (1/s)

Sponge layer width in cells

Print status every N seconds of simulation time (0 = default every 100 steps)

Simulation end time, interpreted in time_unit (default s).

INITIAL layer-thickness profile for the ocean path (sim_type = 'ocean'); ignored by the coastal regimes. Distinct from vcoord_type, which selects the RUNNING coordinate: this knob only decides what h_layer is seeded to at t=0.

Constituent amplitudes (m)

Constituent angular frequencies (rad/s)

Constituent phases (radians)

Unit applied to the long-time fields: t_end, status_interval (logging cadence), ocean_diag_dt_out (diag-write cadence). dt_fixed and dt_max stay in seconds — see their docstrings. Read at namelist parse, then multiplied through so the rest of the code keeps using seconds internally. Valid values: "s", "min", "hr", "day", "year". Year is 365.25 days (Julian, matches MOM6 / UDUNITS convention). Default "s" preserves backward compatibility — every existing nml is bit-identical.

Dedicate one MPI rank per node as I/O server (MPI only)

Enable coupled hydrostatic vertical layers (alternative to NH)

Vertical coordinate type: “sigma”, “zsigma”, “zstar”, etc.

Surface wind stress in x-direction (Pa)

Surface wind stress in y-direction (Pa)

z_fixed_profile = "list": nominal layer thicknesses (m), SURFACE FIRST (MOM6 ALE_COORDINATE_CONFIG = "PARAM:..." / vgrid order). Exactly nz_layers leading positive entries; the rest unset (<= 0, default -1). The profile’s total is its sum; a column deeper than that puts the excess in the bed layer.

z_fixed_profile = "tanh": surface-layer nominal thickness (m). Must satisfy nz_layers*z_fixed_dz_top < max_depth.

Nominal layer-thickness profile of vcoord_type = "z_fixed" and of vcoord_type = "zstar" (MOM6 z*, which dilates this same nominal profile per column by (H + η)/H): "uniform" (default — max_depth/nz_layers everywhere, byte-identical), "list" (the thicknesses in z_fixed_dz, surface first) or "tanh" (a hyperbolic-tangent stretching from z_fixed_dz_top at the surface, scaled to sum to &ocean_topo_nml max_depth; see rdb_vcoord :: z_fixed_nominal_dz). The z_fixed target builder, the closed-face mask, k_top, the cavity partial-top rule and the bed partial-cell rule all read the profile.

z_fixed_profile = "tanh": transition centre as a fraction of the layer-index span, [0, 1] (0 = surface, 1 = bed).

z_fixed_profile = "tanh": transition width as a fraction of the layer-index span (> 0; small = abrupt).

Partial-step z-level face closure under vcoord_type = "z_fixed" (Adcroft, Hill & Marshall 1997; Losch 2008 for the ice-shelf cavity). A layer whose nominal geopotential range lies inside the bed — or inside the ice draft — carries an inert FILLER of thickness zstar_h_min (<= H_VANISHED). A velocity face where layer k is a filler on EITHER side is not a thin passage, it is a WALL for that layer: no normal velocity, no mass / tracer flux, and FREE-SLIP on the tangential component. Leaving it open makes the FV pressure gradient integrate across a staircase step of height Δz_step, which drives |ρ′|·g·Δz_step/(ρ₀·dx) out of a resting stratified state — independent of the filler thickness, so no h-gate reaches it.

Vanishing-layer floor (m). Layers that would land below the local bed get clipped to this thickness rather than going to zero.

Target physical thickness of the surface layer (m). 0 => auto: fall back to the lite (uniform sigma) profile per column (effectively zstar-lite per-column). Set > 0 to anchor the surface layer at a fixed thickness regardless of H.

Number of “fine” near-surface layers using stretching. 0 => auto (use max(1, nz/3))

Surface-concentration stretching: “log” | “uniform”

Added after scale, once at read.

Path to the pre-regridded field(x, y, time) NetCDF. Blank ⇒ tag inactive.

Multiplied into the slab once at read (unit conversion).

Variable name inside file. Required when file is set — a blank var with a non-blank file is a fail-loud setup abort, never a guessed name.

Exterior barotropic u, east (m/s).

“legacy” (default) = v1 Flather (u_ext=0, no interior vel); “full” = half-characteristic form (Flather 1976).

Zero the T-cell wet-mask in the ghost cells beyond every solid WALL edge at setup, so the derived C-grid face masks (wet_u/wet_v/wet_q) are 0 at the wall face and the existing per-stage mask_layer_velocities clears the wall-normal velocity (MOM6’s mask-in-the-update; the halo beyond a wall is land). DEFAULT ON (2026-07): a solid wall must carry zero normal velocity; leaving it garbage (flux-masked but nonzero) was a bug. Set .false. to recover the pre-fix legacy answer for a closed-basin run (e.g. to reproduce an old double_gyre baseline exactly). Without it a flat all-wet channel leaves the wall-normal face velocity as unmasked garbage — the wall flux is masked so mass conserves, but the raw v/u at the wall drifts to the maxvel clamp and paints a spurious vorticity band in diagnostics. Only WALL edges are touched: periodic edges keep their wrapped (wet) ghosts, open/OBC edges keep the interior value (OBC override). Default .false. ⇒ wall ghosts untouched ⇒ bit-identical to legacy runs (this also changes the continuity PPM mirror-h at the wall, which is why it must be opt-in).

Exterior barotropic v, north (m/s).

Inflow nudging timescale (s, Marchesiello et al. 2001). 0 = off.

Outflow nudging timescale (s). 0 = off.

Apply the 18.6-yr nodal factor f_c + equilibrium/nodal phase (V_c + u_c) to the open-boundary tidal elevation forcing, keeping the interior body tide and the boundary tide phase-consistent (same generator, same shared &ocean_tides_nml reference epoch). Default .false. ⇒ legacy static-phase OBC sum (bit-identical). When .true. the *_tidal_phase becomes a Greenwich phase LAG (subtracted).

Running-mean weight (0=full running mean, 1=instant; default=1).

Clamp on the Orlanski nondimensional phase speed (grid cells / step).

“anomaly” (default) = v1 BT-mean + zero-gradient anomaly. “orlanski” = per-layer implicit-upwind radiation (Orlanski 1976).

Inflow reservoir length scale (m). 0 = instantaneous inflow (T_data applied on inflow unconditionally — same as v1 for inflow direction when feature is active).

Outflow reservoir length scale (m). 0 = feature disabled (v1 sign-switch path unchanged). Non-zero engages the implicit Marchesiello et al. 2001 reservoir on all open-ish edges.

Exterior barotropic v, south (m/s).

When .true., extend the sponge relaxation to h_layer + tracers. Default .false. — preserves bit-identity of existing sponge runs.

Peak relaxation rate (1/s) at the outer face of the band.

Number of cells in the sponge band. 0 disables the sponge.

BC type string for the west edge. Recognised: “wall” (default), “open”, “tidal”, “clamped”, “sponge”, “chapman”, “periodic”. “tripolar_fold” is accepted on the NORTH edge only (Murray bipolar cap; requires grid_config=”tripolar” + periodic w/e).

Exterior barotropic u, west (m/s).

Minimum effective BBL thickness (m). Typical 0.1 m.

Multiplier on the bed-layer (k=1) drag tendency only. Default 1.0 = bit-identical.

Background velocity floor (m/s) for the distributed quadratic form. Typical 0.1 m/s.

Quadratic drag coefficient (dimensionless). Typical 2.5e-3. Zero disables the quadratic branch.

Side-wall drag coefficient (dimensionless) for the channel drag. Zero (default) disables the branch.

Lateral side-wall (channel) drag: a per-layer Rayleigh sink at every layer with a partially blocked cross-stream perimeter. Default .false.. Flat-bottom/all-wet ⇒ exact no-op.

Bottom-drag variant: “quadratic” (default, log-layer du/dt = -C_d·|U|·u/h) or “linear” (Rayleigh du/dt = -r·u).

Bottom-boundary-layer thickness (m) over which drag is distributed. Zero (default) = bed-only mode.

Backward-Euler (implicit) bottom drag: u^{n+1}=u/(1+dt·λ), unconditionally stable. Default .false. = explicit forward-Euler (bit-identical) but unstable on thin shelf layers. Recommend .true. for shallow-shelf runs.

Linear Rayleigh coefficient (1/s). Active when form = "linear".

If .true., derive n_inner at setup from the gravity-wave CFL, evaluated per WET cell (local depth with local cell size, MOM6 set_dtbt; land never limits it). Default .false..

MOM6 split (BT_force + eta_PF): the barotropic substep is forced by the depth mean of the FULL slow PGF — baroclinic part included — minus only the free-surface term that PGF itself carries at the η it was evaluated on (none for the surface-relative MONT/FV_LITE/FV_WRIGHT forms), which the substep’s own -g_bt·∇η replaces. .false. restores the legacy split, which subtracted the WHOLE depth-mean PGF and so never let the barotropic mode feel the baroclinic bottom-pressure gradient (JEBAR; ~0 Sv through Drake Passage on the global 1° case against MOM6’s ~160 Sv).

BT continuity-flux velocity projection weight (MOM6 BEBT). Default 0.1 = MOM6’s own default for that parameter: damps the barotropic gravity waves at |λ|² = 1 − b·a² per substep (a = c·dt_bt·k_eff), which is what damps the barotropic grid-scale mode under pred_corr. 0.0 = pure forward-backward Euler (neutral, the pre-2026-09-22 default). The FB stability limit tightens to a ≤ 2/√(1+2·bebt) (MOM6 set_dtbt’s 1+2·BEBT factor).

Wide-halo march-in width. -1 (default) = AUTO, which resolves to 0: the march-in is OPT-IN. A decomposed run with the march-in is not bit-identical to the serial run over variable bathymetry or with open boundaries (measured by test_ocean_decomp_bitid_mpi; flat-bottom closed / periodic / spherical cases are), so the default keeps every decomposition bit-reproducible. 8 (BT_HALO_AUTO_WIDTH) is the recommended explicit width. 0 = explicit off (per-substep grouped exchange, v1 bit-identical). Even positive value: widen the BT ghost band to nghost + bt_halo and fire one grouped exchange every bt_halo/2 substeps. Odd values are rounded DOWN to even with a warning. Fail-loud exclusions (apply when bt_halo is set EXPLICITLY > 0; AUTO instead silently resolves to 0): wet/dry enable, use_cont_type, upstream_h_face, tides enable, psurf enable, porous enable, cavity_dyn enable, supergrid/tripolar grid_config. The set of record is bt_halo_auto_exclusion; keep it and the validate_config checks in lockstep.

PR-2 (bt-rem-from-av-rem): bt_rem_u/v built from the SAME viscous remnant the layered momentum solve uses, instead of the linear-piston substep_drag law or the static 1.0 no-op. Matches MOM6’s barotropic solver: av_rem = Σ_k frhat_k·visc_rem_k (the visc_rem depth mean, frhat_k the face layer fraction derive_bt_from_layers already uses), then bt_rem = mask·av_rem**(1/n_inner) (zero where mask·av_rem <= 0), built ONCE per barotropic call — after the visc_rem producer, before the inner substeps — into the existing bt_work%bt_rem_u/v multiplier the substep loop already reads. This is the fix for the bbl_glue day-253 instability (the barotropic solver was seeing weak explicit drag while the layers were strongly glued): the fast mode now feels the SAME friction. Default .false. ⇒ no answer change. Self-sufficient (D1 follow-up) — the visc_rem producer runs whenever this is on, decoupled from the retired correction_visc_rem; mutually exclusive with substep_drag (bed drag would be double-counted — once inside visc_rem via the glue/implicit_drag fold, once again via the linear piston) and with bt_halo > 0 (the wide-halo BT clone’s metrics_w/halo-widened arrays carry no av_rem/visc_rem ghost width yet — same posture as porous). Composes with wave_drag (multiplied in after).

Safety fraction on the shallow-water CFL bound when auto_n_inner = .true. (typical 0.65-0.7).

When .true. (and use_cont_type), the η-correction bound uses the BT_cont face-by-face flux limits. Default .false..

Adds a per-layer baroclinic-PGF retro-correction for the η change during the BT substep. Requires pgf%form = "fv_mom6".

RETIRED (2026-10-02) — setting it .true. is a fail-loud validate_config error. It distributed the barotropic increment by h_face(k)/⟨h⟩_h, which is not energy-conserving: beyond the barotropic ΔKE it adds a positive-definite source ½Δ²·H·(κ−1), κ = Σh³Σh/(Σh²)², plus a shear feedback that grew the stretched-z_fixed 1-degree Southern Ocean to a non-finite state on day 16. MOM6 has no h-weighted fold (its barotropic acceleration is the same in every layer). The key stays registered only so the refusal can say why; drag-aware weighting is correction_visc_rem.

RETIRED (2026-10, D1 follow-up) — setting .true. is a fail-loud validate_config error. This used to weight the BT-corrector fold by visc_rem(k)/⟨visc_rem⟩_h instead of uniformly, but MOM6’s barotropic solver gives every layer the SAME u_accel_bt via accel_layer_u — no visc_rem weight — folded into up BEFORE vertvisc distributes it via the SAME implicit friction the BBL glue uses, so MOM6 never damps it twice. This fold did, and the SECOND, unbounded visc_rem_k/⟨visc_rem⟩_h ratio is what NaNs the 1-degree Southern Ocean z* open-step case under bbl_glue at step ~40 (see validate_config’s refusal message for the measured isolation). visc_rem_chain is the replacement and uses the UNIFORM fold, matching MOM6. The key stays registered only so the refusal can say why; the underlying kernel dispatch (apply_bt_correction’s use_visc_rem) and its direct unit tests are untouched.

MOM6 wt_u parity for the BT FORCING assembly: weight each layer’s contribution to F_bt_u/v (and to the PGF-projection subtraction, which must use the same weights) by h_face·visc_rem(k) instead of h_face. Layers the implicit friction will immediately damp — grounded sliver stacks under the vdiff BBL glue — then contribute nothing to the fast loop, which otherwise integrates the spurious grounded-layer PGF’s depth-mean ballistically (PGF_BUG.md §9). Self-sufficient (D1 follow-up) — the visc_rem producer runs whenever this is on, decoupled from the retired correction_visc_rem.

rdb_barotropic_coupling::frhat_h_face_step’s per-layer face-thickness closure for EVERY barotropic depth mean that reads a layer’s face thickness: derive_bt_from_layers, face_depth_mean_u/v, face_depth_mean_rem_u/v (forcing_visc_rem’s wt_u), apply_bt_correction’s open/visc_rem folds, and — through face_depth_mean_u/v — compute_bt_rem_from_visc_rem’s av_rem.

Mode-split barotropic substeps per outer step. 0 (default) routes the unsplit ocean_dyn_step; >= 1 routes the split solver with that many fast substeps (production ~30-100). Overwritten when auto_n_inner = .true..

pred_corr predictor fraction (MOM6 BE, 0.6 in the control run): the predictor’s provisional velocity advances to BE·dt; only the corrector takes the full step. Unused under ssp_rk2.

MOM6 continuity-inversion parity (SPEC §4 S2b): pass visc_rem_u/v into the slow-continuity transport-matching renormaliser, switching it to the γ-weighted du + Jacobian form (u_cor = u + du·γ_k) — a heavily-frictioned layer receives a smaller share of the barotropic correction than an undamped one, and the same weighting lands in the mass fluxes. Behaviour change when on (the flux expression tree gains the γ factors). Self-sufficient (D1 follow-up) — the visc_rem producer runs whenever this is on, decoupled from the retired correction_visc_rem.

MOM6 RESCALE_STRONG_DRAG: under strong_drag, bt_rem**n_inner /= av_rem exactly (the rational form is only an approximation), so the barotropic-correction Δu is rescaled by min(bt_rem**n_inner/av_rem, 1.0) before it is folded into the layers, keeping the correction consistent with the TRUE depth-mean remnant. Requires strong_drag = .true.; inert (and refused) otherwise.

Outer time-integration scheme for the split-explicit ocean path (SPEC §4 S3/S4). Both schemes are supported and both are under test; they differ in what they cost you.

MOM6 BT_STRONG_DRAG (default .false.): replace the plain power form with the rational approximation bt_rem = mask·n_inner·av_rem/(1 + (n_inner-1)·av_rem), which damps LESS aggressively per substep for a given av_rem — recommended only if the plain av_rem**(1/n_inner) form still leaves the acceptance-gate run unstable (D3, a measured deviation to report, not a default). Requires bt_rem_from_visc_rem = .true.; inert otherwise (checked in validate_config).

Multiplies the per-face BT velocity update by a damping factor every inner step.

When .true. (default, bit-identical) the BT fast loop integrates its LIVE relative vorticity + KE gradient ((ζ_bt+f)·v − ∇KE_bt) every substep. .false. = MOM6 parity: the fast loop carries planetary Coriolis only (f·v_at_u, live velocities); ζ_bt-advection and ∇KE stay FROZEN inside the depth-mean forcing F_bt_*_fast, exactly like MOM6’s q = f/D weights (btstep_find_Cor — planetary only, no ζ, no KE term). The subtract_fast_cor_ref reference reduces to its f·v̄ part so the τ=0 cancellation stays exact. Motivation: the live nonlinear terms host an exponential ~4Δx wall-trapped BT mode on shelf rims (wall-biased KE stencil: wall faces contribute zero KE, so −∇KE points into walls and corners) that the bound_kh-clamped viscosity (λ·dt ≤ bound_coef/8) cannot damp at large dt — the 1024²×100 dt=600 SE-corner column evacuation (h-guard step 106).

When .true., the BT chain uses per-face upstream-PPM column-sum thickness instead of centred h_face. Default .false..

When .true., the BT substep uses a piecewise-cubic flux-bounded closure instead of uh = u·h_face. Default .false..

PR-3 (visc_rem audit + unification, D1 — revised 2026-10 once MOM6 settled the BT-correction fold question): ONE switch for exactly MOM6’s vertvisc_remnant/av_rem/bt_rem set — DEFAULT ON since PR-4 (the flip, 2026-10), TOGETHER with &ocean_vdiff_nml hvel_mom6/bbl_glue — MOM6 gives the barotropic solver the same friction the layered vertical solve applies; without it the BBL glue’s bed piston never reaches the fast mode (NaN at day 253 on the 1-degree Southern Ocean closed-face case under glue-only). .false. restores the pre-PR-4 linear-piston substep_drag behaviour for bt_rem (mutually exclusive with this switch, D2 below) — equivalent to switching on the visc_rem PRODUCER (decoupled from the retired correction_visc_rem, runs whenever any consumer below needs it) plus forcing_visc_rem (MOM6 wt_u), renorm_visc_rem (MOM6 continuity u_cor = u + du*visc_rem, which IS how MOM6 ties visc_rem to a velocity correction) and bt_rem_from_visc_rem (MOM6’s barotropic av_rem/bt_rem viscous-remnant depth mean) all at once. The barotropic-correction fold (apply_bt_correction) stays UNIFORM under this switch — MOM6’s accel_layer_u never weights it by visc_rem, so neither does this chain; see correction_visc_rem’s docstring for why that fold is retired, not folded in here. Same caveat as before: without &ocean_vdiff_nml implicit_drag=.true. or bbl_glue=.true., visc_rem stays ≡ 1 and the whole chain is a legal, warned, no-op. The three *_visc_rem consumer knobs (forcing_visc_rem/renorm_visc_rem/bt_rem_from_visc_rem) stay individually settable (never retired) for the existing fine-grained tests, each an equivalent SUBSET of this switch, never a superset, and each now SELF-SUFFICIENT (no longer “requires” a separate producer knob — the producer runs whenever any one of them is on). strong_drag/ rescale_strong_drag stay separate namelist keys per D1 (MOM6 has its own BT_STRONG_DRAG/RESCALE_STRONG_DRAG parameters for them) and require this switch (or bt_rem_from_visc_rem) on, same as before. accel_visc_rem (&ocean_vdiff_nml) is likewise NOT part of this chain and is RETIRED (see its own docstring) — no MOM6 state-update equivalent. Default .false. ⇒ no answer change (PR-4 is the default flip). D2: mutually exclusive with substep_drag (checked via bt_rem_from_visc_rem’s existing requirement, which this switch satisfies identically to setting it directly).

Master switch for the barotropic linear (Rayleigh) wave drag — the bulk energy sink for the barotropic tide (Egbert & Ray 2001; Jayne & St Laurent 2001), MOM6 BT_LINEAR_WAVE_DRAG. Default .false. => bit-identical. Composes multiplicatively with substep_drag inside bt_rem_u/v.

Reserved for PR-14 (MOM6 BT_WAVE_DRAG_FILE); unused today.

r_H filler: “uniform” (global scalar), “roughness_proxy” (resolved-bathymetry-variance proxy for Jayne & St Laurent <h^2>), or “file” (reserved for PR-14’s NetCDF map reader — fail-loud not-implemented today).

Ceiling on the resolved-bathymetry <h^2> proxy (m^2; ⇔ h_rms <= 158 m) for wave_drag_form = "roughness_proxy".

Topographic wavenumber kappa (1/m) for wave_drag_form = "roughness_proxy". Same default as &ocean_tidal_mixing_nml kappa_itides — keep them equal.

Reference near-bottom buoyancy frequency N_bot (1/s) for wave_drag_form = "roughness_proxy".

Piston velocity r_H (m/s) for wave_drag_form = "uniform".

Global tuning multiplier on r_H (MOM6 BT_WAVE_DRAG_SCALE).

Reserved for PR-14 (MOM6 BT_WAVE_DRAG_VAR); unused today.

Analytic draft shape. "none" (default): z_draft = 0 everywhere — the identity, even with enable = .true.. "flat": uniform draft_depth inside the shelf box [draft_x0, draft_x1] x [draft_y0, draft_y1], 0 outside (the open ocean beyond the calving front at draft_x1). "linear": z_draft = draft_depth + draft_slope*(x - draft_x0) inside the same box, clipped at 0 below. "file": a static 2-D NetCDF draft — NOT implemented (fails loud); the MPI-correct static-2-D reader is a later slice, and it is the only route to an ISOMIP+ draft, which has no analytic form (Asay-Davis et al. 2016 §3.1.1).

Draft amplitude (m, positive down) — the uniform value for "flat", the value at draft_x0 for "linear". Under draft_source = "thickness" it is an ice THICKNESS instead.

draft_config="file": path to the NetCDF carrying the static ice draft. The variable named by draft_var must be rank 3 in FORTRAN storage order (x, y, t) — which is how a C or Python writer (and ncdump) spells (nTime, ny, nx) — with a time coordinate variable; RECORD 1 is read and the field is never re-read. There is NO horizontal interpolation: the file must already be on the model grid (nx x ny physical cells), exactly as bathymetry_file and &ocean_zinit_nml file require. Single rank only (fail-loud otherwise).

draft_config="file": the SIGN CONVENTION of the file values. There is no default that guesses from the data — the two conventions differ by the whole ice load, and a field that is partly open water (zeros) is indistinguishable by inspection.

"linear" only: d(draft)/dx, dimensionless (m of draft per m of x). Positive deepens the ice base toward +x. Converted from metres to GRID units at the dispatch (metres on a Cartesian grid, degrees on spherical/curvilinear), the same way &ocean_topo_nml slope_scale is.

What the draft is prescribed FROM. "draft" (default): the geometry above IS the ice-base depth. "thickness": the formula gives an ice THICKNESS, converted by the Boussinesq-isostatic (flotation) relation z_draft = rho_ice*h_ice/rho_0. "in_situ" (true isostasy, p_ice = g*integral(rho_hat)) needs a per-column root find, does not admit exact discrete rest, and is deliberately NOT implemented (fails loud).

draft_config="file": name of the 2-D variable to read. The default is the ISOMIP+ geometry file’s own spelling (Asay-Davis et al. 2016 Sect. 3.3).

Western edge of the shelf box (m, GLOBAL physical coordinate), and the ANCHOR of the "linear" profile (draft_depth is the draft AT draft_x0), which is why "linear" requires a finite value here. The default +/-1e30 on all four bounds is the “no limit on this side” sentinel: the shelf then covers the whole domain INCLUDING the ghost band, which is what a shelf that reaches a wall needs (a box stopping at x = 0 puts a phantom calving front one cell outside the west wall).

Eastern edge of the shelf box = the CALVING FRONT (m): beyond it the draft is 0 (open ocean). Default: no eastern limit.

Southern edge of the shelf box (m). Default: no limit.

Northern edge of the shelf box (m). Default: no limit.

Master switch. Requires the ocean multilayer path, the split solver, &ocean_pgf_nml form="fv_mom6", vcoord_type in {sigma, zstar} and a single rank; mutually exclusive with wet/dry, porous barriers, sea ice, bt_halo > 0, tidal SAL and gfs_scale /= 1 (every one of those fails loud at configure, naming the knob and the reason).

Sanity bound: if more than this fraction of the interior columns ground, configure fails loud rather than silently running a domain that is mostly land.

GROUNDING cutoff (m): a column whose water thickness b - z_draft is below this is LAND — it goes through the same seed_wet_mask_impl the bathymetry uses, so the static metric-zeroing land mask and the finite land-state hold for free. Never a thin film of water under grounded ice. ISOMIP+ §3.1.5 leaves the choice to the modeller and notes ~40 m (two cells) for z-level models; sigma is less restricted, hence 10 m.

Ice density (kg/m^3), consulted ONLY by draft_source = "thickness".

Trim the INITIAL column under the ice so it is at rest (MOM6 TRIM_IC_FOR_P_SURF, trim_for_ice). The load p_ice_ref = rho_ref*g*z_draft is the displaced weight at the REFERENCE density; a stratified column’s displaced water weighs g*int_{-z_draft}^{0} rho dz, and the difference is a depth- uniform bottom-pressure gradient the MOM6 barotropic split (&ocean_bt_nml bc_pgf_forcing) adjusts to. With the knob on, the load is kept (the ice MASS is what is prescribed) and each loaded column’s initial top is moved to the depth s where g*int_{-s}^{0} rho dz = p_ice_ref, i.e. an initial eta = z_draft - s (a few cm under ISOMIP+ COLD), with T/S then evaluated at the trimmed layer centres. Closed form, exact at the discrete FV interfaces: requires &ocean_eos_nml eos="linear" and &ocean_zinit_nml enable, source="linear" (the analytic profile is what defines rho above the ice base); anything else fails loud. Default off => bit-identical.

Top drag coefficient C_D,top entering the MELT friction velocity u*^2 = C_D (U^2 + u_tide^2). ISOMIP+ Table 4 p. 2483. The least constrained number in the subject: the literature spans 1.5e-3 (Holland & Jenkins 1999 Table 1) to 9.7e-3 (Jenkins et al. 2010 Table 2). This knob does not yet drive any momentum drag — top drag is Phase 4; here it only scales u* for the exchange velocities.

Master switch. Requires &ocean_cavity_dyn_nml enable, &ocean_eos_nml tfreeze_set="isomip" and &ocean_forcing_nml enable_components; mutually exclusive with atmospheric surface forcing (wind stress, surface restoring, shortwave penetration, the uniform scalar q_heat/q_salt) until the per-cell cover mask lands, and with sea ice. Every one of those fails loud at configure, naming the knob and the follow-up. &ocean_psurf_nml composes freely: the liquidus reads the ASSEMBLED ms%p_top = p_ice_ref + sf%p_surf, so an atmospheric load under the shelf depresses the freezing point with no extra wiring.

Turbulent exchange-velocity law. "const_gamma" (default) is gamma = Gamma*u* — Jenkins, Nicholls & Corr (2010) eqs. (1),(2),(5) p. 2300 and the ISOMIP+ form. "hj99" is Holland & Jenkins (1999) eqs. (14)-(18) p. 1792 (needs a non-zero Coriolis parameter under the cover). "yung25" is Yung et al. (2025) “StratFeedback” eqs. (7)-(8) p. 5832. Every other name the kernel’s enum reserves (jenkins91, rosevear22, vt19, mk18, burchard22, jenkins21) PARSES but is refused at configure naming CAVITY_MELT_NOT_IMPLEMENTED — a reserved law and a typo must stay distinguishable.

Thickness (m) below the ice base over which the far-field (T, S, u, v) are thickness-averaged, with a partial last layer. Metres, deliberately, never “layer nz”. The melt rate is roughly linear in the thermal driving it is handed, and how far from the ice that was sampled is the dominant resolution artefact in the subject (Gwyther et al. 2020; Burchard et al. (2022) Table 2 p. 15 — the all-bulk error GROWS under refinement; Yung et al. (2026) p. 2074). No protocol prescribes a value; 10 m is this repository’s default and it must be held FIXED across any vertical-coordinate comparison, or the comparison measures the sampling depth instead.

How the meltwater reaches the ocean.

Dimensionless salt-transfer coefficient Gamma_S. Negative = unset ⇒ resolved to gamma_t/35, the ISOMIP+ ratio (Asay-Davis et al. (2016) Table 4 p. 2483, after Jenkins, Nicholls & Corr (2010) p. 2309: the ratio “should lie somewhere in the range 35-70. Adopting a value at the lower end of this range…”). Zero and positive values are taken literally, so gamma_s = 0 is refused as a range error rather than silently re-triggering the default.

Dimensionless heat-transfer coefficient Gamma_T of gamma_t = Gamma_T*u*. ISOMIP+ starting guess, Asay-Davis et al. (2016) §3.2.1 p. 2487 — a starting guess, not a constant of nature: the protocol has participants tune it, and Yung et al. (2026) Table 2 p. 2058 shows the twelve submissions spanning 0.011 to 0.2. Re-derive it per vertical coordinate; a single value across a coordinate sweep makes the sweep measure its own tuning.

Ice-side heat conduction. "insulating" (default) is q_ice = 0, which the ISOMIP+ protocol PRESCRIBES (Asay-Davis et al. (2016) Table 4 p. 2483 sets kappa_i = 0 and p. 2485 instructs participants not to use the H&J99 advection-diffusion scheme); t_ice is then unread. "adv_diff" is Holland & Jenkins (1999) eq. (31) p. 1794 in its melting asymptote, which collapses to q_ice = m_mass*c_i*(T_b - T_ice) and is zero on freezing. "diffusive" is RESERVED and refused: it changes the melt/freeze BRANCH logic, not just a coefficient.

Ice salinity (g/kg), >= 0. Zero is the ISOMIP+ value (Asay-Davis et al. (2016) Table 4 p. 2483). It must stay strictly below the far-field salinity — that inequality is what the three-equation root bracketing rests on — and a column violating it is COUNTED and given zero melt, not guessed at.

Ice interior temperature (degC), read by ice_conduction="adv_diff" only — Holland & Jenkins (1999) Table 1 p. 1790 uses T_S ~ -25. Under "insulating" the kernel substitutes exactly zero, so a stale value here cannot leak into an insulating run.

RMS tidal velocity (m/s) in the melt friction velocity — ISOMIP+ Table 4 p. 2483 and eq. (27) p. 2485, after Jenkins, Nicholls & Corr (2010) eq. (10) p. 2309. TRAP, and the protocol says it outright (p. 2486): “The computation of top and bottom drag do not incorporate utidal” — it belongs to the melt u* ONLY.

Friction-velocity floor (m/s) — Yung et al. (2025) eq. (14) p. 5836, value from their Table 2 p. 5838. It exists because “a friction velocity of zero (perhaps created by initialising the model at rest) will result in identically zero melt … which would be inconsistent with the presence of heat available for melting”.

What to do with the volume freshwater="mass" adds to a CLOSED domain. Requires freshwater="mass".

Open-boundary condition config (&ocean_bc_nml). All defaults = wall — existing nmls without this block are bit-identical to prior behaviour.

Static ice-shelf cavity geometry (&ocean_cavity_dyn_nml). Default-OFF ⇒ bit-identical.

Ice-shelf basal-melt thermodynamics (&ocean_cavity_melt_nml). Default-OFF ⇒ bit-identical.

Shared time-varying NetCDF input reader (&ocean_data_nml, PR-14). Two knobs only — see ocean_data_config_t.

File-backed surface forcing (&ocean_dataovr_nml, PR-15) — the first production consumer of data. Default-OFF ⇒ bit-identical.

Forensic probes (&ocean_debug_nml). Default all-off ⇒ bit-identical.

Surface-flux component-set gate (&ocean_forcing_nml, PR-12). Default off ⇒ byte-identical.

Along-coordinate tracer Laplacian (&ocean_hdiff_nml). Default kappa_h = 0.0 ⇒ bit-identical.

Sea-ice slot config (&ocean_ice_nml). Default off ⇒ byte-identical (PR 0 scaffold).

Sea-ice initial-condition config (&ocean_ice_ic_nml, PR 24). Default conc_config="zero" ⇒ byte-identical.

Lagrangian grounding-stability controls (&ocean_isopycnal_nml). Default-OFF ⇒ bit-identical.

Multi-rank MPI debug / tuning controls (&ocean_mpi_nml). Default off ⇒ bit-identical.

Ice-shelf TOP drag (&ocean_tdrag_nml). Default off ⇒ bit-identical.

Dynamic wetting/drying (&ocean_wetdry_nml). Default off ⇒ byte-identical.

Floor used by the PPM positivity limiter (MOM6 GV%Angstrom_H). When ppm_limit_pos = .true. and a cell’s centred thickness is at or below this floor, the PPM face reconstruction is forced to a constant (= upwind for that cell); above the floor the parabola’s interior minimum is forced to h_min if it would dip below. Default 1e-6 m preserves “pure positivity” semantics. MOM6 typically uses 1–4 m (≈ 2·Angstrom_H). No-op when ppm_limit_pos = .false..

Positive-definite continuity (MOM6-prevention + Roundabout-conservation). When .true. the split layer continuity guarantees every layer stays >= h_lim after each direction pass — the PPM reconstruction edges are floored at 2·h_lim (P1, MOM6’s positive-definite PPM) and (P2, later) the per-donor outfluxes are scaled down (never thickness-inflated ⇒ zero mass created). h_lim is derived at setup: angstrom_h on VCOORD_LAGRANGIAN, else 0 (⇒ the floor is inert on non-Lagrangian coords). Default .false. ⇒ untaken branches only ⇒ bit-identical. Fail-loud composed with &ocean_wetdry_nml enable (that module owns its own barotropic limiter; composition deferred).

MOM6 PPM_limit_pos analogue. When .true., the PPM face-thickness reconstruction in continuity adds a positivity-preserving limiter that prevents the parabola from dipping below h_min inside a cell. At thin / vanishing layers (shelf-break, seamount, wet/dry boundary) the reconstruction collapses to a constant — equivalent to upwind for that cell. Bounds mass flux through near-vanishing layers by the actual layer thickness.

Continuous flux model in the uhbt/vhbt Newton renormalisation (see continuity_t%renorm_consistent_flux): a layer whose upwind donor flips under the barotropic correction carries (u0 + du)·h_face(new donor) instead of the historical flux0 + du·h_face(new donor), which jumps by u0·(h_new − h_old) at the flip and leaves the solve with no root when uhbt falls in the gap — a wrong-sign layer transport, an O(η) mismatch between the layer and barotropic free surfaces at every thickness jump (a sigma layer over a bathymetric step), and an exponentially pumped barotropic grid-scale mode under pred_corr (finding B of the vertical-coordinate stability matrix). Newton is bracketed by bisection (MOM6 zonal_flux_adjust). Default .true. (MOM6 behaviour, maintainer decision 2026-09-22); it is bit-identical to the historical model on every face where no donor flips. .false. restores the historical discontinuous model. Ignored by the wet/dry single-step form.

MOM6 vol_CFL analogue. When .false. (default) the continuity-PPM face thickness is the downwind PPM EDGE value (the CFL→0 limit), bit-identical to the pre-knob behaviour. When .true. the donor-side face thickness is the swept-volume integral of the reconstructed parabola (adds the missing O(CFL) term). Fixes the dt-sensitive near-bed mass residual at steep shelf breaks.

Master switch. Requires vmix%use_closure + thermo%enable_thermodynamics (validated at configure).

Convective tracer diffusivity (MOM6 KD_CONV), m^2/s.

Trigger threshold on N^2 (MOM6 BV_SQR_CONV), s^-2.

Kv_conv = prandtl_conv * kd_conv (MOM6 PRANDTL_CONV), nondim.

MOM6 BOUND_CORIOLIS: clamp the energy-scheme Coriolis acceleration into the range of the four neighbouring VELOCITY-form estimates (f+ζ)·v BEFORE the KE-gradient subtraction. Kills the thin-layer PV blow-up: q=(f+ζ)/h_corner is huge at a vanishing edge and q·vh can inject an acceleration the local velocity field cannot support; the (f+ζ)·v bound caps it. Energy scheme only (form="sadourny_energy", fail-loud otherwise) — MOM6’s double-gyre control run enables it. Default .false. ⇒ untaken branch ⇒ bit-identical.

PV corner-thickness construction for the energy scheme. "cell_mean" (default, bit-identical) — Roundabout’s wet-area-weighted 4-cell mean h_corner = Σ(wet·area·h)/Σ(wet·area), floored at H_MIN_PV. "mom6_area" — MOM6’s exact form: q = abs_vort·Area_q/(hArea_q + vol_neglect) with hArea_q = Σ(mask·area·h), Area_q = Σ(mask·area). The two are algebraically IDENTICAL above the floor (same numerator + denominator); they differ ONLY in the vanishing-thickness guard — Roundabout CAPS q at abs_vort/H_MIN_PV, MOM6’s vol_neglect is pure 1/0 armor (no cap), so at sub-H_MIN_PV corners MOM6’s q is LARGER. For realistic layers (h ≳ 1e-12) the knob is a round-off no-op. Energy scheme only (fail-loud otherwise).

Coriolis-advection variant: “sadourny” (default, enstrophy form), “sadourny_energy” (energy-conserving transport form q·vh), “sadourny_hk” (Hollingsworth-Källén).

PV face-interpolation scheme, orthogonal to form (Sadourny path only): “centered” (default, 2-point corner average ⇒ bit-identical) or “weno3”/”weno5”/”weno7” — upwind-biased WENO-Z reconstruction of the corner absolute vorticity onto the velocity faces (MOM6 WENOVI{3,5,7}TH), sharpening submesoscale PV fronts without the centred form’s global dissipation. weno5/weno7 (radius-3/4 stencils) require nghost >= 3/4 (fail-loud otherwise).

MOM6 mass-consistent CorAdCalc: the mom6 corrector’s Coriolis-adv consumes the predictor continuity’s renormalised transports (ms%mass_flux_*_layer — the uh/vh from the same solve that produced u_av) instead of recomputing u·h_face from the evaluation-state prognostics. Requires form="sadourny_energy" + &ocean_bt_nml split_scheme="pred_corr" (fail-loud). Default .false. = bit-identical.

Size of the reader’s field registry. A registration beyond this is a fail-loud abort (max_fields >= 1 is enforced by the schema min=1).

Log every bracket advance (field, record pair, weight, model time) via logger%info. Default off — quiet per-step path.

Climatology period (s). Required > 0 when time_mode = "cyclic"; ignored otherwise.

Master switch. Default off ⇒ no registration, bit-identical. Requires RDB_ENABLE_NETCDF=ON at build time.

Evaporative mass flux (kg/m^2/s, <= 0) -> surface_flux%evap.

Net surface heat flux (W/m^2, positive down) -> surface_flux%heat_added.

Liquid precipitation (kg/m^2/s, >= 0) -> surface_flux%lprec.

.true. clamps a query outside the file time axis to the end record (one warning); default .false. aborts.

Surface salt flux (kg salt/m^2/s, positive salinifies) -> surface_flux%salt_flux.

Added to the model time before the file-axis lookup (s) — shifts the model epoch onto the file’s epoch.

Zonal wind stress (Pa) -> surface_stress%tau_x (east faces).

Meridional wind stress (Pa) -> surface_stress%tau_y (north faces).

Shared time-axis mode for every active tag: linear (interannual; out-of-range is an abort unless oor_clamp), cyclic (climatology wrapped through cycle_period), or static (record 1, read once). Per-tag overrides are a documented v2 gap, not an oversight.

In-cell reconstruction for the conservative diagnostic vertical remap (z_fixed / density vgrids): “pcm”, “plm”, “ppm” (default), “ppm_h4”, or “pqm”. All conserve the column integral; PPM is the accurate default (PCM stair-steps). Inert when every diagnostic is on the native “layer” grid.

Unified diagnostic selection list — one authoritative spec that modifies the canonical default set. Whitespace/comma-separated entries, each name[:attr]... with self-identifying colon attributes (order-free):

Cadence (s) at which the diag manager fires every variable’s time op (INSTANT writes, MEAN flushes the accumulator, etc.).

Enable per-step diag-manager hook in driver_run_ocean.

Output file basename — output_rank_filename appends the per-rank suffix. Final path: <output_dir>/<filename>_rank_NNNNNN.nc.

When .true., remapped (non-layer) diagnostics fill target cells that overlap no water (below-bottom / pinched-out in a shallow column) with a missing sentinel and tag the NetCDF variable with _FillValue / missing_value. Default .false. => those cells read 0 (bit-identical to the legacy writer).

Number of entries used in rho_levels (0 = none).

Number of entries used in sigma_levels (0 = auto-uniform).

Number of entries used in z_levels (0 = none).

Number of entries used in zstar_levels (0 = auto-uniform).

Element width of the DIAGNOSTIC NetCDF data variables: “double” (default => byte-identical to the pre-knob writer) or “single” (fp32 => ~half the bytes per frame; diagnostic output is write-bandwidth bound, so this is the main lever on its cost). fp32 carries ~7 decimal digits — ~1e-5 degC, ~1e-5 PSU, ~1e-9 m/s at 1 m/s — orders of magnitude below the model’s own discretisation error.

console-conservation totals (Mass/KE/Salt/Heat + sea-ice area) and the salt/heat closed-budget out/src terms use order-invariant extended-fixed-point (EFP) summation (rdb_efp + halo_allreduce_efp_list) instead of plain FP !$acc parallel loop reduction(+:acc) + MPI_SUM — the printed console is identical on every rank count (1 included) and reduction order, and the Error residual is formed via efp_real_diff (a fixed-point difference) rather than a double subtraction of two already-quantised totals. It runs only at the status cadence and changes diagnostic TEXT only, never the trajectory. .false. restores the pre-PR-32 FP console (whose last digits depend on the decomposition). See docs/CAPABILITIES_AND_LIMITATIONS.md’s conservation-contract section for the achievable guarantee + the EFP_MAX_RANKS = 131072 envelope.

Output target potential-density bin edges (kg/m^3, strictly increasing, light->dense) when vgrid = “density”. No auto-fill — validate_config requires n_rho_levels > 0 whenever the density vgrid is selected (globally or per-diagnostic).

Output sigma levels — cumulative fractions (0..1, shallow->deep) when sigma output is used. Empty => auto-generate nz_ml uniform fractions.

Default output vertical grid for layer-shaped diags: “layer” (native nz_ml layers, default), “z_fixed” (conservative remap to z_levels), “sigma” (terrain-following, sigma_levels), “zstar” (SSH-tracking, zstar_levels), or “density” (isopycnal bins, rho_levels — requires n_rho_levels > 0, strictly increasing). Per-diagnostic overrides via the diags list :layer/z/sigma/zstar/density attribute.

Output z-levels (m, positive downward) when vgrid = “z_fixed”.

Output z reference interface depths (m, positive-down, shallow->deep; deepest = H_ref) when z output is used. Empty => auto-generate nz_ml uniform depths (== sigma; supply a non-uniform reference for z* to differ).

Equation-of-state variant: “linear” (default, two-tracer), “wright” (Wright 1997 rational), “roquet_spv” (Roquet et al. 2015 specific-volume polynomial). “roquet_spv” is incompatible with the “fv_wright” PGF (fails loud at configure). “teos10” parses but is refused at configure (not implemented; Roquet SpV IS the 75-term TEOS-10 polynomial fit).

Reference pressure (Pa, >= 0) at which the model’s POTENTIAL density ms%rho_layer is evaluated. Default 0 (surface density, σ₀) ⇒ bit-identical to every run before this knob existed. Read by the “wright” and “roquet_spv” variants; “linear” has no pressure dependence and ignores it.

Named seawater freezing-point (liquidus) coefficient set for eos_freezing_point, which evaluates the linear form

OM4 Ekman coefficient (MOM6 MSTAR2_COEF2).

“add” / “max” vs the interior closure’s kv/kt (MOM6 EPBL_IS_ADDITIVE).

Rotational mixing-length rolloff (MOM6 EKMAN_SCALE_COEF).

Master switch. Requires vmix%use_closure; replaces the KPP overlay (configure logs the override).

Stokes SL-average depth fraction (MOM6 LA_DEPTH_RATIO).

Enhancement coefficient (MOM6 LT_ENHANCE_COEF).

Langmuir-number exponent (MOM6 LT_ENHANCE_EXP).

Stability-modified La coefficients (MOM6 LT_MOD_LAC1..5); set all to 0 for the unmodified La.

Cap on the multiplicative enhancement.

“rescale” (multiplicative) / “additive” (MOM6 EPBL_LANGMUIR_SCHEME).

Mixing-length floor, m (MOM6 EPBL_MIN_MIX_LEN).

Shape-function exponent (MOM6 MIX_LEN_EXPONENT).

Bisection instead of false position (MOM6 EPBL_MLD_BISECTION).

Self-consistent MLD root-find (MOM6 USE_MLD_ITERATION).

Max MLD iterations (MOM6 EPBL_MLD_MAX_ITS).

MLD convergence tolerance, m (MOM6 EPBL_MLD_TOLERANCE).

Seed from the previous step’s MLD (MOM6 MLD_ITERATION_GUESS).

Constant-scheme mstar (MOM6 MSTAR).

Cap for OM4/RH18; off when < 0 (MOM6 MSTAR_CAP).

OM4 stabilizing coefficient (MOM6 MSTAR2_COEF1).

Convective mstar reduction in [0,1] (MOM6 MSTAR_CONV_ADJ).

“constant” / “om4” / “rh18” (MOM6 EPBL_MSTAR_SCHEME).

Convective PE -> TKE efficiency (MOM6 NSTAR).

Earth rotation rate, 1/s (MOM6 OMEGA).

Blend |f| with 2*Omega (MOM6 ML_OMEGA_FRAC).

Kv = prandtl*Kd into the momentum solve (MOM6 EPBL_PRANDTL).

RH18 mstar fits (MOM6 RH18_MSTAR_CN1..CS2).

Ekman-depth / TKE-decay-scale ratio (MOM6 TKE_DECAY).

Compute the per-column TKE budget diagnostic terms.

Transition-layer shape floor, must be in [0,1) when iterating (MOM6 EPBL_TRANSITION_SCALE).

Langmuir-turbulence enhancement (LF17 wind-only path, no wave model needed). MOM6 EPBL_LT / USE_LA_LI2016.

“cube_root” / “rh18” (MOM6 EPBL_VEL_SCALE_SCHEME).

kappa in Kd = vstarkappamixlen (MOM6 VON_KARMAN_CONST).

Overall vstar multiplier (MOM6 EPBL_VEL_SCALE_FACTOR).

RH18 mechanical surface vstar factor (MOM6 VSTAR_SURF_FAC).

Convective weight in the velocity scale (MOM6 WSTAR_USTAR_COEF).

Bodner mechanical (u) weight in w’u’ = (mstar·u³+nstar·w*³)^{2/3}.

Bodner convective (w*) weight in the same w’u’ estimate.

FK08 coefficient Ce (0.06-0.08).

Bodner (2023) efficiency coefficient Cr (nondim). Default 0 (MOM6 default) ⇒ no Bodner transport until set (~0.02-0.08).

Master switch (off => bit-identity).

|f| regularisation floor (1/s).

Floor on w’u’ (m²/s²) — pure 1/0 armour when u* and the surface buoyancy flux both vanish.

Running-mean MLD filter time-scale (s) used when the MLD is retreating. 0 (default) => filter off => instantaneous EPBL MLD (bit-identical). Positive damps step-to-step MLD swings.

B2 res_fn double-counting hook (hard error if on without B2).

mu cubic-tail extension (smoother); default 0 = exact mu.

Bodner et al. (2023) frontogenesis-arrest MLE variant: replaces the Ce/|f| timescale with Cr·Δs·|f|·h/w'u', the frontal-arrest length set by boundary-layer turbulence. Overrides ce/ use_mom_mixrate. Default off ⇒ classic Fox-Kemper.

FK11 momentum-mixrate timescale vs bare Ce/|f|. Default off.

Master switch. Default .false..

Scalar constant bottom heat flux (W/m^2, positive up). Typical ~0.05-0.1.

Master switch. Default off ⇒ gm_compute_transports no-ops. Requires &ocean_slopes_nml enable (configure-time error).

Thickness diffusivity KhTh, m^2/s (constant-fill for v1; production 1e2-1e3). 0 ⇒ no bolus transport even when enabled.

Fraction of the diffusive CFL the face KH may use.

Slope magnitude above which the safe-streamfunction blend takes over (MOM6 slope_max).

Units of the Cartesian domain extent (grid_config="cartesian"), mirroring MOM6 AXIS_UNITS: “meters” (default), “degrees”, “km”. With “degrees”/”km” the domain is sized by len_lon/len_lat below and &grid_nml dx/dy are DERIVED (ignored as inputs). “meters” ⇒ &grid_nml dx/dy used verbatim (bit-identical).

Coriolis source for the metrics f-fill: “beta_plane” (default, f_0 + betay) or “planetary” (2omega*sin(geolat)).

Grid generator: “cartesian” (default, uniform), “spherical” (lon-lat sector), “supergrid” (MOM6 mosaic NetCDF), “tripolar”.

South edge of the physical domain (degrees) — spherical only.

Total y-extent of the Cartesian domain in axis_units (MOM6 LENLAT). Paired with len_lon; > 0 ⇒ dy = to_metres(len_lat)/ny. NOTE: spherical cell sizes reuse &grid_nml dx/dy as dlon/dlat in DEGREES when grid_config="spherical".

Total x-extent of the Cartesian domain in axis_units (MOM6 LENLON). <= 0 (default) ⇒ unset. > 0 ⇒ a uniform dx = to_metres(len_lon)/nx is derived, where for “degrees” to_metres(v) = rad_earth·v·π/180 (arc length, no cos(lat) — MOM6 cartesian) and for “km” to_metres(v) = 1000·v.

Longitude (degrees) of the first tripolar cap pole; partner at lon_pole + 180.

West edge of the physical domain (degrees) — spherical only.

Planetary rotation rate (rad/s) — used by “planetary”.

Join latitude (degrees) for “tripolar”: lon-lat below, bipolar Arctic cap above.

Earth radius for the spherical metric (m).

Path to the MOM6 supergrid NetCDF — “supergrid” only.

Constant horizontal tracer diffusivity (m^2/s). Zero (default) is a no-op — the kernel short-circuits without touching hTr, so this knob is bit-identical when unset. Stability bound (explicit forward-Euler), checked at configure using the coarsest dx/dy (Cartesian metres): kappa_h * dt_therm * (1/dx^2 + 1/dy^2) <= 0.5

Background harmonic viscosity (m²/s) for the lateral closure. Negative (default) = derive from nu_h and kh_vel_scale.

Hard cap on the per-face harmonic viscosity (m²/s).

Anisotropy direction vector (n1,n2) in grid i,j components. Default (1,0) = grid-i. Normalised internally.

Anisotropy-direction mode. Only mode 0 (grid-relative (n1,n2) = aniso_dir) is implemented; other values rejected at configure (no silent fall-back).

CFL safety coefficient for the per-cell viscosity limiters (harmonic when stress_tensor, biharmonic always when nu_4 > 0 or Smag_AH). Default 0.8; must be ≤ 1.

MOM6 BOUND_KH analogue for the velocity-Laplacian paths: per-face clamp of the harmonic viscosity to bound_coef·0.125/(dt·(1/dx²+1/dy²)) (~¼ of the forward- Euler stability limit). Prevents an over-large nu_h (or flow-aware coefficient) from putting the frozen depth-mean viscous forcing on the barotropic mode into the phase- reversed ANTI-damping regime at grid scale, which pumps an exponential rim-trapped barotropic mode via the split-RK2 Δu corrector (the 600² Lagrangian double-gyre blow-up). Recommended .true. for any config where nu·dt·(1/dx²+1/dy²) ≳ 0.1. Default .false. ⇒ bit-identical.

Dimensionless Leith coefficient. Typical range 1.0–2.0.

Nondimensional biharmonic Leith constant. Drives lateral_closure="leith_biharm" (per-face nu4_face = C_lb · grid_sp⁶ · inv_PI6 · |∇²ζ|, clamped to [nu_4_bg, nu_4_max]). Default 0.0 = no-op (startup warning if leith_biharm selected at 0.0).

Dimensionless Smagorinsky coefficient. Default 0.15 (range 0.1–0.2).

Anisotropic Laplacian viscosity magnitude (m²/s; Smith & McWilliams 2003). When positive AND stress_tensor = .true., adds direction-tensor terms aligned with aniso_dir. Default 0 ⇒ isotropic, bit-identical. Stress-tensor path only.

Velocity scale (m/s) for the grid-resolution viscosity floor nu_min = kh_vel_scale · dx. Typical 3e-3 m/s.

Live velocity-scale viscosity coefficient (m/s). When positive, a state-dependent A_vel = kh_vel_scale_live·dx·|u| is max-combined into the per-face harmonic viscosity every step. Default 0 ⇒ off, bit-identical.

Lateral-mixing closure tag. “none” (default) keeps the scalar nu_h. “leith” = Leith vorticity-gradient scaling; “smagorinsky”/”smag” = Smagorinsky strain-rate scaling. Either gives a per-face viscosity floored at ah_bg, capped at ah_max.

Lateral coast BC (static land mask). .false. (default) = FREE-SLIP (corner vorticity/shear strain multiplied by wet_q); .true. = NO-SLIP (factor 2 - wet_q). All-wet ⇒ bit-identical.

Constant biharmonic eddy viscosity for momentum (m⁴/s). Default 0.

Background biharmonic viscosity floor (m⁴/s) for the flow-aware path.

Static hard cap on per-face biharmonic viscosity (m⁴/s). The forward-Euler biharmonic-CFL bound is enforced automatically per cell (scaled by bound_coef), so this is just an extra ceiling.

Constant horizontal eddy viscosity for momentum (m²/s). Default 0 = bit-identical; wind-driven runs need this set.

Resolution-scaled Laplacian viscosity (Hallberg 2013). When .true. the dynamic Laplacian viscosity is multiplied by the VarMix resolution function Res_fn ∈ [0,1] before the ah_max clamp (biharmonic nu_4 never scaled). Requires &ocean_varmix_nml enable=.true.. Default .false. ⇒ bit-identical.

Smagorinsky biharmonic. When .true. the biharmonic viscosity becomes flow-aware nu4_face = C_b · L⁴ · |D|, clamped to [nu_4_bg, nu_4_max]; the scalar nu_4 is bypassed.

Nondimensional biharmonic Smagorinsky constant. Typical 0.015–0.06.

Horizontal-viscosity operator. .false. (default) = velocity-Laplacian A·∇²u with global caps, bit-identical. .true. = thickness-weighted stress-divergence form (momentum-conserving) with per-cell CFL limiter and coast masking; reduces to the Laplacian on uniform-grid/h/all-wet.

Linear-EOS reference salinity (PSU) — the S at which the haline anomaly term vanishes.

Linear-EOS reference temperature (°C) — the T at which the thermal anomaly term vanishes.

Linear-EOS thermal-expansion coefficient, DIMENSIONAL (kg/m³ per °C) — see beta_S for the conversion from the fractional 1/°C coefficient most protocols quote.

Linear-EOS haline contraction coefficient, DIMENSIONAL (kg/m³ per PSU).

Reference temperature (°C) for the Eady IC at z=0, y=y_mid.

Meridional gradient of T (°C/m) for the Eady IC.

Vertical stratification (°C/m) for the Eady IC.

Amplitude (°C) of the random temperature perturbation seeded into the Eady IC to break symmetry.

RNG seed for the Eady IC perturbation.

Geostrophic-adjustment IC: amplitude (m) of the Gaussian SSH bump.

Geostrophic-adjustment IC: e-folding scale (m) of the Gaussian SSH bump.

x-coordinate (m) of the bump centre. Negative ⇒ auto (basin midpoint).

y-coordinate (m) of the bump centre. Negative ⇒ auto (basin midpoint).

IC overlay tag. “” (default) keeps the analytical T(z) + zero-velocity IC. “eady” runs the Eady-front overlay; “geostrophic_adjustment” drops a Gaussian SSH bump on an f-plane; “baroclinic_jet” runs the two-layer reduced-gravity baroclinic-instability jet (SIM_DETAILS.md §5).

Baroclinic-jet IC: interface displacement amplitude Δξ (m). Spec “tuned” default 200 m; faithful uses 100 m.

Baroclinic-jet IC (ic_config="baroclinic_jet"): tanh jet half-width L (m). Spec “tuned” default 40 km; the Julia- faithful case uses 100 km. Read only for the baroclinic-jet IC.

Per-layer initial density (kg/m³), k=1 bed → k=nz surface, mirroring MOM6’s COORD_CONFIG="gprime" IC. Any positive value bypasses the EOS-derived init and writes the user-specified densities straight into ms%rho_layer at the end of ocean_state_seed_from_cfg. Count of non-sentinel (>= 0) entries must equal nz_layers. Default -1.0 sentinel ⇒ existing init path (bit-identical).

Baroclinic-jet IC: front-localised meander amplitude as a fraction of Δξ (A_pert = pert_amp_frac·Δξ). Spec default 0.2.

Baroclinic-jet IC: zonal perturbation wavenumber (integer number of wavelengths across the periodic domain). Integer so the analytic x-continuation wraps exactly (k_x·L_x = 2π·pert_nx). Spec default 3.

Reference density (kg/m³) for the linear EOS and Boussinesq PGF — the density at (T_ref, S_ref).

Linear density-range IC — MOM6 COORD_CONFIG="linear" analogue. When >= 0, generates nz_layers linearly-spaced layer densities spanning [rho_lightest, rho_lightest + rho_range] and writes them into ms%rho_layer (bed k=1 heaviest → surface k=nz lightest), reusing the layer_rho_init write path — so the column scales by just bumping nz_layers, no hand-listed densities. Mutually exclusive with layer_rho_init. Default -1.0 sentinel ⇒ off (bit-identical).

Linear density-range IC: total top-to-bottom density contrast (kg/m³), MOM6 DENSITY_RANGE (default 2.0). Read only when rho_lightest >= 0.

Baroclinic-jet IC: rest thickness of the upper (surface, k=2) layer H₁ (m). The lower (bed, k=1) rest thickness is max_depth − upper_layer_rest. Spec §4 default 500 m (H₂ = 1500 m over the 2000 m column).

PR 62: weight the atmospheric stress AND the ice-ocean drag in the EVP momentum balance by the face ice concentration a_u = 0.5*(ci(i-1,j) + ci(i,j)) (same expression/edge convention as ice_ocean_stress_flux_impl’s a_u), giving the textbook m du/dt = grad.sigma + a*(tau_a - tau_w) (Hibler 1979 eq. 1) and an EXACTLY closing ice<->ocean momentum budget at every fractional cover, not just steady free drift. Both terms must be weighted together – weighting the wind alone converts today’s leak (zero at steady free drift) into a permanent one. Default off => byte-identical.

Advective sub-iterations per transport call (SIS2 NSTEPS_ADV, default 1). dt_adv = dt_therm / adv_substeps.

Prescribed slab-atmosphere air temperature (degC) for the v1 restoring atmospheric-forcing filler. SF(T) = restore_lambda*(T - air_temp).

SIS2 ICE_STRENGTH_CSTAR — ice-strength exponent constant [nondim].

SIS2 ICE_CDRAG_WATER — ice-ocean drag coefficient [nondim].

SIS2 CFL_TRUNCATE (SIS2 default 0.5) — transport-CFL ceiling on the FINAL ice velocity the EVP call hands to transport: |u| <= 0.95*cfl_trunc*areaT(donor)/(dt_transport* dy_cu). 0 (Roundabout default, house bit-identity rule) disables the clip entirely => byte-identical. Recommended production value 0.5 (“instability can occur past 0.5”, SIS2).